Riverlane Demonstrates Data-Driven DEM Augmentation to Reduce Surface-Code Logical Errors
quantumcomputingreport.com Nov 8, 2026

Riverlane Demonstrates Data-Driven DEM Augmentation to Reduce Surface-Code Logical Errors

AI-summarised brief · reviewed before publication

Riverlane has unveiled a data‑driven technique that expands standard detector error models (DEMs) with correlated error mechanisms extracted directly from syndrome measurements in surface‑code quantum processors. By limiting candidates to combinations of existing DEM elements, the method assigns clear logical operators and integrates into fast graph‑based matching decoders. Numerical simulations of planar surface codes with SI1000 noise and added two‑qubit XX crosstalk show a logical error‑rate reduction of up to 75 % at distance 15, corresponding to an error‑suppression factor Λ of 2.06 versus 1.71 for the unaugmented model. On Google’s below‑threshold superconducting hardware the augmented priors improve Λ to 2.02, a 9 % gain, and cut the distance‑7 logical error rate by 14 %. The approach needs no extra calibration and scales.

💡 Why It Matters

  • · Turning hidden crosstalk into decoder priors accelerates progress toward fault‑tolerant quantum computing without modifying existing hardware.